24 tháng 4 2013

How to have an Effective Digital Analytics Strategy? - From Chandler Nguyen's blog

Bài gốc từ đây http://www.chandlernguyen.com/2013/04/effective-digital-analytics-strategy.html


First of all, if you want to have a completely solid, effective analytics strategy, please read Avinash Kaushik’s blog or read his book “Web Analytics an hour a day”. Everything I ‘ve learnt about Web Analytics, I learnt from reading his blog & his book. From his blog, you should be able to find a list of other suitable blogs at the right hand corner as well.
When it comes to Digital Analytics Strategy, what he recommended (and i found that’s true) was that to follow the 90/10 rule.

90/10 rule

Only 10% of your time and effort should be spent on choosing the right tool. 90% of the company’s time & money should be spent on getting the right analysts in place. Any analytics tool is just (Surprised) a TOOL, no more no less and we still need human to use the tool.

An Analyst job is not easy. Of course, he/she needs to be able to create pretty graphs, or skillful at creating pivot tables to communicate the points across. However, those are just the basics and more often than not we could find many junior people who could do that stuff. The more (ONLY) important thing is the ability to recommend actionable business insights based on the company strategy, marketplace, competition so on and so forth. To do so, the analyst needs to have a broader understanding of the whole business operation, not only from the online marketing side, but the overall marketing strategy, sales performance, customer services etc…
For simplicity sake, in Vietnam context, Google Analytics would be powerful enough for the majority of companies. There will be some exceptions of course but their percentage is small. To put things into perspective, business reasons are something i consider while recommending Google Analytics as well. Of many free tools, Google Analytics seems to be most frequently used in Vietnam so familiarity is there. For a country where English is still a foreign language, having an interface in Vietnamese does help a lot.
Enterprise tool like Omniture, WebTrends etc… would cost quite a bit in Vietnam context and the time/human resource taken to implement these tools is substantial. After having a paid Enterprise tool, the company would need to send some staff for training as well, which may not be that productive since people tend to move quickly from one company to the next in Vietnam. Maintenance cost is something you need to bear in mind as well.

Asking the right business questions

Yup it should be the second thing after getting the right analysts for the job.
Always start with business questions, trying to understand what it is that the company is trying to do with the existing/new website/microsite/marketing campaign. Answers like “Increasing brand awareness, more sales” are too generic in nature and we probably need to dig deeper.
We wouldn’t recommend starting with looking straight into some standard reports/dashboards yet because they may fill with jargons you don’t understand or don’t find valuable. Knowing the average page per visit alone doesn’t help sales manager one bit. Or knowing average pageview per visit increases 20% over the past 1 month alone doesn’t help you make any kind of decision. Does that mean our brand awareness increase 20% over the past 1 month? No, a page view by itself is just a page view.
We suffer from too much data, data in silo with no meaning, no actionable insights after reading them rather than the lack of data so please do NOT start looking straight at some sample dashboards. Take your time and think deeper about what business issues you are trying to solve running this digital campaign or building this website?
  • For example, who your target audience is? What do they often do online? How do we know if the right target audience is visiting our site?
  • What do we want visitors to know about our brand when they are on the site? How do we know if visitors “read/understand” the message that we want them to?
  • Could visitors find what they are looking for easily on your site?
  • How do online visit affect offline store traffic? How do we know which marketing initiative is working better than the other?
  • Which product line is working better on the website?
  • How can we increase our prospective sales lead database using the website?
  • Could customers who want to order something from us do so easily? Where do we put our hotline? Our menu?
  • Can we increase our overall revenue by 10% in the next 3 months using online channels?
After listing down some business questions that you want to answer, let put yourself in the shoes of the visitors and visit your company website/microsite.
I know you probably visit your company website/microsite hundreds of time but let’s do it at least one more time.
What do you see on the homepage? How do you feel? What do you understand about the company by visiting the homepage?
If the business objective is a new product launch, what do you understand from reading the product page? Do you even want to read it in the first place? Do you see the sign up button? Have you tried doing so?
The purpose of this post is not about going into user experience s
The next step is to translate these business questions into key performance indicator or KPIs.

Business Objectives to SMART KPI

You can not optimize what you can’t measure. So after listing down all of the business questions that you want to answer, it’s time to create SMART KPIs.
The KPIs need to be related as directly as you can to the business questions you have.
For example:
  • If your target audience is in Ho Chi Minh city, then one of the very first report you should look will be geo location report to see how many visitors are from Ho Chi Minh city.
One of the KPIs can be how many unique visitors from Ho Chi Minh visit the site monthly?
How many new visitors from Ho Chi Minh do you acquire monthly?
Then you should go deeper in different traffic sources to understand which traffic source/campaign drive visitor from Hanoi or Hue to your website? So that you can consider pausing them if they are paid media.
  • If you launch a new car and you want people to register to test drive it, then you may want to track which traffic source brings the highest number of sign ups.
You KPIs could be:
    • How many unique visitors see the landing page with the new car?
    • The cost per new visitor to the landing page?
    • Which traffic source is performing based on new visitors coming to the landing page and not bouncing back right the way?
    • How many per cent signs up? The cost per sign up?
  • If you have 2 important messages that you want to bring across to your audience then it may be worth to look at first the loading speed of your content page, then to see how many unique visitors actually “read” your two important pages? How do they come to those pages and where do they go after these page? These reports are readily available on Google Analytics.
Your KPIs could be:
  • Loading speed of the special landing page
  • The number of unique visitors who visit the landing page
  • Bounce rate on the special landing page
  • Time spent on the landing page
  • If you expect visitors to be curious and take some actions after reading the page, then you could have that as one of the KPI as well, even actions like looking at the page with the location of your business.

Data authenticity

First you need to make sure that you understand how the analytics tool measures your KPIs.
It should be pretty straight forwards and standard with Google Analytics and some of the bigger tools. However each tool may define things differently.
Properly implement the tracking code:
  • Tag all pages on your website
  • Put tracking codes at the bottom of the page (or at the portion of the page, there has been so much discussion on this but for simplicity sake I think just go with either).
  • Tracking code should be inline. They shouldn’t be inside some fancy tables or frames etc…
  • If you have multiple sub domain, consider different “profiles” carefully
  • If there is a cross domain issue i.e. users are redirect to another domain while they take some action on your site like e commerce function, please set up cross domain tracking properly
  • Define and properly set up conversion tracking/goals settings
Use code to track onclick function if you want to.
  • Be aware of redirects of any kind because they will mess up your traffic source analysis.
  • Test and re-test the data to make sure what you see on the report is actually correct. Please test as many scenarios as possible. For example, you may want to test when someone search for a keyword on Google, see the site on the Search Engine Result Page, click on it and take some actions. The report should properly record the keyword that user uses, date of the visit, the action taken.
Test with email marketing, banners, with directly enter the site. Test booking and see if the revenue data is being captured correctly, the items in the shopping basket are being recorded.
The more time you spend testing now, the better it will be later on because wrong data is worse than no data.
  • If your site uses flash, special codes need to be generated to track actions embedded in flash file. The same goes with Video.
  • Consider the default cookie length in case you have special need
  • Less than 10% difference is ok. No measurement is perfect so when you test or look at the report, if the difference between the numbers from two different tools or from two different sources is less than 10%, try to live with it. You may want it to be perfect but hey, in the offline world, 10% accuracy is gold and i doubt 10% is the case in Vietnam. We should be looking at larger than 10% discrepancy for offline reports from newspaper circulation to reach.

Digital Analytics responsibility should belong to Marketing/Sales/Brand team

Well after you have everything properly set up and tested out, it’s time to return the login details to the marketing team/sales team from IT team.
Many companies have the IT department send out monthly standard website statistics to the marketing team. To me it’s a waste of resources doing so because:
  • IT team does not know about all of the marketing initiatives happening so they would not be in the best position to pull relevant reports
  • To analyze you have to zoom into each channel, segmentation like crazy and running custom report. Standard reporting won’t do you much good
So I am a big believer in training the Marketing team to be able to use the interface, navigating around and find information they need on their own.

One dashboard for each level

Each position/level in a company has different need for information/data. Director level would be more interested in top line data, revenue data or the like. However, marketing executive doing SEO would be interested in keyword ranking, traffic from brand terms vs non brand terms etc…
So having separated dashboards for different levels/functions is of paramount importance. It will hopefully solve the issue of NOT reading the reports. Come on after you spend so much time with the reports, people do not even read them??? How could it be?
Well i know it’s sad but it’s true and it’s happening all the time.
People often read reports when their boss asks them some data that they don’t know, when they have to authorize a payment or something of the sort.
To read more about this, please refer to my post about 100 pages report.

Segmentation is the key

This idea is not new. It basically means you don’t look at the aggregate number and try to make decision or next step because normally you can’t. What needs to be done is to dig deeper, separate each channel, each traffic source and examine them one by one.
The behavior of visitors in Ho Chi Minh may be very different from Hanoi for your website even so it may be good to analyze them separately.
For example, if you know the overall website bounce rate is 50%, there’s nothing much you could do. However, if you know that the bounce rate from paid search, to a specific landing page is 50% and there is a call to action on that page, now you are on so something. You may want to change the keyword, ad text or landing page to ensure consistency.

Trend is what to look out for

In my humble opinion, trend is vitally important. Sometimes trend is even more important than the actual value at a particular point in time. They provide the top line view, which is crucial in deriving actions.
When you look at a report, different time frames will give you very different pictures/insights. Try different time unit like daily report, weekly, monthly report as well.
For example, the all time cost per sales is at $2 and your campaign has been running for 1 year. However for the past 3 months, your cost per sales has been going up and now it is at $3.50 (nearly double the all time value)!
When you want to understand trends, to avoid signal noise and to derive the insights faster, sometimes it helps by choosing different time unit that the graph is using like instead of looking at the visits per day, try to at it per week.
Also many tools allow you to do “Compare to past” which is comparing two different time frames in the same graph. This is a powerful function if used correctly because you want to compare apple to apple, taking into consideration the seasonal factors and all.
It is often better to draw two related parameters in one graph as well. For example, I would look at the number of clicks delivered daily and the average cost per click to see if things are really going well for Paid Search. If the line for daily number of clicks going up and average cost per click going down then we are in good position.

Benchmarking is king

Context is everything. A number in silo means little and it is hard to derive what the next step should be.
You could do internal benchmarking or external benchmarking, which is entirely up to you.
External benchmarking may be extremely hard because it’s not easy to find reliable third party data to benchmark against.
For example, if your report says your sales revenue increased 10% over the last month, that may make you feel great. However, data from the industry shows that your nearest competitors increased their sales by 50% last month then your 10% may not be that “great” any more.
Internal benchmarking could be done by comparing different time frame, segmenting and looking at individual component over time, comparing different product lines/microsites/sub domains etc…

That’s all from me for today. Do you agree/disagree? Do you want to add in something?

Không có nhận xét nào: